The Attenuation Paradox and Internal Consistency,
JOHNS HOPKINS UNIV BALTIMORE MD DEPT OF STATISTICS
Pagination or Media Count:
The paper indicates why a concept of true score naturally leads to a belief that test validity must increase with an increase in test andor average item reliability, and why for the classical single factor model first introduced by Spearman is correct. A statistical model used by Loevinger is introduced to establish the attenuation paradox and to explain why this paradox holds in this particular model. It is hoped that mental test methods can include more critical appraisal of commonly used techniques and concepts, as well as a check of methods of test construction against the statistical models.